Messages in this thread | | | From | "Liu, Chuansheng" <> | Subject | RE: [PATCH] firmware loader: Fix the race FW_STATUS_DONE is followed by class_timeout | Date | Thu, 8 Nov 2012 01:31:33 +0000 |
| |
> -----Original Message----- > From: Ming Lei [mailto:ming.lei@canonical.com] > Sent: Wednesday, November 07, 2012 6:40 PM > To: Liu, Chuansheng > Cc: gregkh@linuxfoundation.org; linux-kernel@vger.kernel.org > Subject: Re: [PATCH] firmware loader: Fix the race FW_STATUS_DONE is > followed by class_timeout > > On Thu, Nov 8, 2012 at 12:45 AM, Chuansheng Liu > <chuansheng.liu@intel.com> wrote: > > > > There is a race condition as below when calling request_firmware(): > > > > CPU1 CPU2 > > write 0 > loading > > mutex_lock(&fw_lock); > > ... > > set_bit FW_STATUS_DONE class_timeout is coming > > set_bit > FW_STATUS_ABORT > > complete_all &completion > > ... > > mutex_unlock(&fw_lock) > > Good catch, but your fix isn't correct. > > > > > In this time, the bit FW_STATUS_DONE and FW_STATUS_ABORT are set, > > and request_firmware() will return failure due to condition in > > _request_firmware_load(): > > > > if (!buf->size || test_bit(FW_STATUS_ABORT, &buf->status)) > > retval = -ENOENT; > > > > But from the above scenerio, it should be a succcessful requesting. > > So we need judge if the FW_STATUS_DONE is set before calling abort > > in timeout function firmware_class_timeout(). > > > > Signed-off-by: liu chuansheng <chuansheng.liu@intel.com> > > --- > > drivers/base/firmware_class.c | 7 +++++++ > > 1 files changed, 7 insertions(+), 0 deletions(-) > > > > diff --git a/drivers/base/firmware_class.c b/drivers/base/firmware_class.c > > index 8945f4e..35fffd8 100644 > > --- a/drivers/base/firmware_class.c > > +++ b/drivers/base/firmware_class.c > > @@ -671,6 +671,13 @@ static void firmware_class_timeout(u_long data) > > { > > struct firmware_priv *fw_priv = (struct firmware_priv *) data; > > > > + mutex_lock(&fw_lock); > > mutex can't be used in interrupt context, so one candidate fix is > to convert the timeout timer into delayed work, and hold the > 'fw_lock' during the work function. > > > + if (test_bit(FW_STATUS_DONE, &(fw_priv->buf->status))) { > > + mutex_unlock(&fw_lock); > > + return; > > + } > > + mutex_unlock(&fw_lock); > > + > > fw_load_abort(fw_priv); > > } > > Also the lock of 'fw_lock' should be held when setting 'FW_STATUS_DONE' > in _request_firmware_load(). > > Could you send out a new patch to fix the race? Thanks your review and proposal, has sent the patch V2 to be reviewed. > > Thanks, > -- > Ming Lei
| |